English

Dr. Top-k: Delegate-Centric Top-k on GPUs

Information Retrieval 2021-09-20 v1 Databases Distributed, Parallel, and Cluster Computing

Abstract

Recent top-kk computation efforts explore the possibility of revising various sorting algorithms to answer top-kk queries on GPUs. These endeavors, unfortunately, perform significantly more work than needed. This paper introduces Dr. Top-k, a Delegate-centric top-kk system on GPUs that can reduce the top-kk workloads significantly. Particularly, it contains three major contributions: First, we introduce a comprehensive design of the delegate-centric concept, including maximum delegate, delegate-based filtering, and β\beta delegate mechanisms to help reduce the workload for top-kk up to more than 99%. Second, due to the difficulty and importance of deriving a proper subrange size, we perform a rigorous theoretical analysis, coupled with thorough experimental validations to identify the desirable subrange size. Third, we introduce four key system optimizations to enable fast multi-GPU top-kk computation. Taken together, this work constantly outperforms the state-of-the-art.

Keywords

Cite

@article{arxiv.2109.08219,
  title  = {Dr. Top-k: Delegate-Centric Top-k on GPUs},
  author = {Anil Gaihre and Da Zheng and Scott Weitze and Lingda Li and Shuaiwen Leon Song and Caiwen Ding and Xiaoye S Li and Hang Liu},
  journal= {arXiv preprint arXiv:2109.08219},
  year   = {2021}
}

Comments

To be published in The International Conference for High Performance Computing, Networking, Storage and Analysis (SC 21)

R2 v1 2026-06-24T06:03:13.699Z